What BMI Actually Measures, and What It Leaves Out

Answer first: it is weight divided by height squared, and nothing else

Body mass index is a person's weight in kilograms divided by their height in meters squared. That is the whole calculation. It uses two measurements, and it produces one number in units of kilograms per meter squared.

Because those are the only two inputs, everything else is outside the number. It carries no information about what the weight is made of — bone, muscle, fat, water, the contents of a digestive tract. It carries no information about where on a body the weight sits. Two people with the same index can have very different bodies.

None of that makes the index useless. It makes it a particular kind of tool: a cheap, repeatable population screen that a clinician, a trial protocol, or an insurance rule can apply to anyone with a scale and a tape measure. Understanding which questions it was built to answer is the difference between reading a threshold correctly and reading it as a verdict.

The bands are a 1990s classification, adopted twice

The categories everyone quotes did not emerge from the index itself. The FDA's own draft guidance on developing weight-reduction drugs states the origin plainly. Based on data relating the index to mortality risk, it says, the World Health Organization in 1995 and the National Institutes of Health in 1998 adopted the weight classifications shown in its Table 1.

That table runs underweight below 18.5, normal weight from 18.5 to 24.9, overweight from 25 to 29.9, obesity class 1 from 30 to 34.9, class 2 from 35 to 39.9, and extreme obesity class 3 above 40. Every figure in that list is a cut point on a continuous measurement. Nothing changes physically at 30 that did not exist at 29.9.

The mortality reasoning behind the bands is also in the guidance, and it is more qualified than the bands look. The relationship between the index and risk of death varies by age, sex, race, and other factors such as smoking status. Generally, the annual incidence of all-cause mortality is lowest between 22.5 and 24.9 and increases from 25 up through above 40. A range where a population's average risk is lowest is a statement about a population, not a target for a person.

The agency that uses it publishes its limitations in the same document

This is the part worth reading twice, because it comes from the regulator rather than from a critic. In the section explaining why the index is used for regulatory purposes, the guidance lists the advantages: it is inexpensive, universally available, easy to calculate, reproducible, and correlates strongly with total body fat in nonelderly adults. It also has a long history of use for regulatory purposes, which matters more than it sounds — decades of trials measured the same way can be compared to each other.

Then, in the very next paragraph, the same document states the limits. The index has several limitations. Although a higher index is strongly associated with increased body fat, the index is not a direct measure of body fat, and it does not inform the distribution of excess body fat. In clinical practice, supplementing it with other anthropometric measures, such as waist circumference, may be appropriate in certain individuals.

Two words in there do the heavy lifting. Not a direct measure means the index infers body fat rather than measuring it. Does not inform the distribution means it cannot tell you where the fat is, which is a separate question from how much of it there is.

The alternatives have limits of their own, which is why the index survives

The obvious response is to measure body fat directly instead. The guidance walks through why trials mostly do not, and the reasons are practical rather than ideological.

Skinfold thickness is operator dependent and has relatively poor reproducibility — two people measuring the same body get different answers. Bioelectrical impedance may vary depending on the hydration status of the individual, which means the same body measured at different times of day can produce different results. Imaging modalities such as dual x-ray absorptiometry or magnetic resonance imaging may provide more precise measures of body fat, but they are expensive and require multiple blinded central readers to run in a trial. And the guidance adds a point that is easy to miss: trial results based on imaging changes may not be generalizable to clinical care, because clinical care does not have a central reading facility.

So the index persists in trials, and not because anyone believes it is the best possible description of a body. It persists because it is the one description that a trial in thousands of people, a prescriber in an exam room, and a plan reviewing a claim can all produce identically.

Why the weight trials still report percent change in body weight

There is a small piece of arithmetic behind this that resolves an apparent contradiction. If the index is weight divided by a fixed height squared, then for an adult whose height is not changing, the percentage change in body weight and the percentage change in the index are the same number. The guidance states exactly that. So an adult trial can use percent change in body weight as its primary endpoint and lose nothing.

Growing bodies break that identity, which is why the semaglutide trial in patients aged 12 and older used percent change in the index as its primary endpoint instead. In a population whose height is still moving, the two quantities genuinely diverge, and the trial measured the one that accounts for growth.

This is the reason a percentage from an adolescent trial and a percentage from an adult trial are not interchangeable even though both carry a percent sign. They are measuring different quantities.

A threshold is an entry rule, not a diagnosis

On the approved labels, the index appears where eligibility is defined. The tirzepatide weight product is indicated for adults with obesity, or adults with overweight in the presence of at least one weight-related comorbid condition. The semaglutide weight product carries the same structure for its weight-reduction indication, alongside separate indications with their own entry requirements.

Read that construction carefully and you can see what it is doing. It is a rule for sorting people into and out of a population, written so that a prescriber can apply it consistently. It is not a claim that a number describes a person's health, and the labels do not describe it as one.

The trials show how far the enrolled population can sit from the edge of that rule. In the tirzepatide weight trials the average index at baseline was 37.4, with 29 percent of patients above 40. A rule that opens at 27 with a qualifying condition admitted a group whose average was ten points higher than that. Somebody who qualifies at the low edge of the rule is at the boundary of the population these results came from, not in the middle of it.

The FDA guidance is explicit that the index is a way to identify patients with obesity or overweight in the clinical setting. Identifying a population and diagnosing an individual are different jobs, and this number was built for the first one.

What to do with the number when you meet it

You will meet it in three places: an intake form deciding whether a program will prescribe, a plan's coverage criteria, and a trial description explaining who was enrolled. In all three it is functioning as a sorting rule.

The useful questions are the same each time. What is the cut point, and what else has to be true alongside it — many rules pair a lower index with a required second condition. Was the number self-reported or measured. And if the context is a trial result, where did the enrolled population actually sit relative to the cut point.

What the number cannot answer is what any particular body is made of. The guidance says so in its own words, and it recommends other measurements alongside it for exactly that reason. A question about composition needs a measurement of composition, which is a different instrument with its own trade-offs.